90
H. K. Palo
• Although SI adopts shared vocabularies and ontologies for a better understanding
of meanings, it still requires maintaining and updating the intrinsic information
from various data sources. It needs to preserve the existing domain knowledge
and supporting data maintenance.
• For better maintenance and utilization of the shared ontologies and vocabularies,
it is essential to develop a periodic and centralized updating mechanism.
• The software facilities and infrastructure relying on the shared ontologies have
still issues related to scalability, an open and well-known issue of SW.
A potential solution to aforesaid issues can be using new ontology models or
restructuring the existing ontologies aligning or integrating these semantically with
various vocabularies using equivalences among their properties and classes [22].
However, a unified semantic model for annotation of IoT data, intelligent integration
of distributed data sources, and the proper management and utilization of suitable
sensors, still poses a challenge to this domain. The fusion and composition of different
data sources, searching and discovery of actuators and data sources, fitting to an
application based on their capabilities, reasoning, and analysis on semantic resources
via visualization tools or reasoners needs further consideration [24, 25]. To realize
the full potential of IoT, “things” need to talk to and understand each other. An
even greater obstacle to attaining this outcome is the different data models employed
within the countless different proprietary implementations.
4 Semantic IoT Versus Machine Learning
There have been many machine learning (ML) algorithms found useful in pattern
recognition, speech, emotion identification, etc. [26–28]. The integration of SIoT
with Machine Learning (ML) algorithms has been explored by many researchers in
the field of pervasive computing, ubiquitous computing, wireless sensor networks,
ambient intelligence, human activity recognition, etc. The state-of-art Artificial
Neural Network (ANN) with a backpropagation algorithm has been explored to detect
human movements such as sitting, walking, running, and for smart home activities
[29, 30]. To process the sensor data in IoT, the trends of Convolution NN and Deep NN
have been utilized effectively [31]. Similarly, various pattern recognition tools such as
the Bayesian networks, Naive Base Classifiers, Support Vector Machines, K-Nearest
Neighbor, Hidden Markov Model have been intelligently utilized for home automation, context-aware search systems, navigation systems in IoT based applications
[32, 33].
Integrating SIoT and machine learning undoubtedly will increase the revenue in
any business or industry. It is essential to choose certain words to describe a certain
set of concepts suitably. The choice of a common vocabulary for SIoT to bridge
the semantic gap existing between machines need to be defined. However, existing
organizations use the concept superficially by not transforming it into a true profitgenerating engine in reality. For example, a fast-food industry wants to introduce AI
H. K. Palo
• Although SI adopts shared vocabularies and ontologies for a better understanding
of meanings, it still requires maintaining and updating the intrinsic information
from various data sources. It needs to preserve the existing domain knowledge
and supporting data maintenance.
• For better maintenance and utilization of the shared ontologies and vocabularies,
it is essential to develop a periodic and centralized updating mechanism.
• The software facilities and infrastructure relying on the shared ontologies have
still issues related to scalability, an open and well-known issue of SW.
A potential solution to aforesaid issues can be using new ontology models or
restructuring the existing ontologies aligning or integrating these semantically with
various vocabularies using equivalences among their properties and classes [22].
However, a unified semantic model for annotation of IoT data, intelligent integration
of distributed data sources, and the proper management and utilization of suitable
sensors, still poses a challenge to this domain. The fusion and composition of different
data sources, searching and discovery of actuators and data sources, fitting to an
application based on their capabilities, reasoning, and analysis on semantic resources
via visualization tools or reasoners needs further consideration [24, 25]. To realize
the full potential of IoT, “things” need to talk to and understand each other. An
even greater obstacle to attaining this outcome is the different data models employed
within the countless different proprietary implementations.
4 Semantic IoT Versus Machine Learning
There have been many machine learning (ML) algorithms found useful in pattern
recognition, speech, emotion identification, etc. [26–28]. The integration of SIoT
with Machine Learning (ML) algorithms has been explored by many researchers in
the field of pervasive computing, ubiquitous computing, wireless sensor networks,
ambient intelligence, human activity recognition, etc. The state-of-art Artificial
Neural Network (ANN) with a backpropagation algorithm has been explored to detect
human movements such as sitting, walking, running, and for smart home activities
[29, 30]. To process the sensor data in IoT, the trends of Convolution NN and Deep NN
have been utilized effectively [31]. Similarly, various pattern recognition tools such as
the Bayesian networks, Naive Base Classifiers, Support Vector Machines, K-Nearest
Neighbor, Hidden Markov Model have been intelligently utilized for home automation, context-aware search systems, navigation systems in IoT based applications
[32, 33].
Integrating SIoT and machine learning undoubtedly will increase the revenue in
any business or industry. It is essential to choose certain words to describe a certain
set of concepts suitably. The choice of a common vocabulary for SIoT to bridge
the semantic gap existing between machines need to be defined. However, existing
organizations use the concept superficially by not transforming it into a true profitgenerating engine in reality. For example, a fast-food industry wants to introduce AI